Compositional modeling reduces crude-analysis time, predicts yields

Solomon M. Jacob, Richard J. Quann, Eugene Sanchez, Maria E. Wells
Mobil Technology Co.
Paulsboro, N.J.

This article about compositional monitoring is the first of three articles that discuss Mobil's lube technologies. The second article will review Mobil's lube processes and commercial examples, and the third article will discuss its catalyst technologies and commercial applications.

Compositional monitoring is a molecular-based method with three components:

• High-detail hydrocarbon analysis (HDHA) to analyze the crude
• Structure-oriented lumping (SOL) to describe the crude mathematically
• Reaction rules to model real-world reactions.

HDHA is a proprietary technique to probe the petroleum's molecular structure and composition. In parallel, Mobil developed a new modeling technology, called SOL, to organize and simulate the complex chemistry of petroleum mixtures at the molecular level.^1-3

Molecular-based models benefits

• Analytical techniques provide some level of detail on molecular composition and molecular structure (for example, HDHA).
• A method exists to represent the molecular structures and the complex reaction chemistry for a very large number of components (for example, SOL).
• Molecular structure-property relationships are developed to calculate the kinetic, thermodynamic, physical, and quality properties (reaction rules).

Compositional modeling can rapidly estimate the value of the crude to Mobil. For each Mobil refinery, the crude-dollar value is assessed. Spot purchases and long-term crude supply contracts can be negotiated without months of expensive product testing. In the past, only a handful of crudes could be approved per year; today there is no limit in the number of approvals.

In the past, with no way of knowing ahead of time how a crude change would affect the base-oil quality, the refinery conducted expensive test runs and performed only limited crude changes. With compositional modeling, it is possible to simulate crude changes ahead of time and specify the process condi tions needed to produce high-quality lubes.

The knowledge of the value of any stream allows Mobil to optimize lubricant production and profits at every step of the manufacturing chain. This product-quality assessment is used to manage the risk of any change in the crude type, process conditions, or base oil.

If the product-quality assessment model indicates a low risk for a particular change, then the change can be implemented with concurrent validation testing. If the assessment is high risk (poor or unknown quality), the change is not approved until a full evaluation is conducted. For some high-risk cases, the change would not be considered further.

For example, one Mobil refinery processed an additional 50,000-70,000 b/d of crude for lube production as a result of compositional modeling. In some cases, it has been possible to process crudes which normally would not have been considered as lube crudes by themselves. In combination, however, the once-rejected crudes can produce high-quality lubes.

Previous model

The traditional process of approving a crude for lube manufacturing used lumping schemes. These schemes were lengthy, trial-and-error processes that involved costly refinery test runs, extensive product testing, and waiting periods of up to 1 year. When approvals were finally issued, they were limited to the specific operating conditions of the test run and subject to formulation restrictions in some product applications.

Lumping partitions the entire molecular population into a small number (about ten) of lumps, based on limited measurements of mixture composition, similar chemistry, similar physical properties, and the model's intended purpose. This approach reduces complexity and eliminates the need for a detailed molecular characterization at the expense of its usefulness.

For example, the model shown in Fig. 1 [135,982 bytes] was developed in the 1970s to predict gasoline yield for the fluid-catalytic cracking (FCC) process. It represents gas oil as eight lumps based on the crude analytical capabilities of that time-broad molecular categories and boiling ranges. Although this model is trained to predict gasoline yield from this limited feedstock characterization, it cannot predict gasoline composition or how feedstock composition impacts the required quality specifications.

Organizing the composition

For example, a vector with a one in the first increment (A6) and a zero for all other increments represents a benzene molecule. Naphthalene is constructed with A6 and A4 increments, having one for the first two increments of the vector and zero for all others, as shown in Fig. 2 [99,140 bytes]. Other examples of SOL representations of molecular structure are shown in Fig. 3 [102,738 bytes]. Molecular stoichiometry and molecular weight are easily computed from the contributions of the incremental stoichiometry.

A complex mixture in SOL is assembled as a set of vectors. Each vector corresponds to a molecule or an ensemble of closely related structural isomers. The composition of the mixture is expressed as the weight or mole percent of each vector.

It is helpful to visualize the mixture using the concepts of molecular class and homologous series. Naphthalene is an example of a molecular class, and its homologous series is composed of all naphthalenes in the mixture. Naphthalenes have varying R (number of carbon atoms on side chains), br (number of side chains), and me (number of methyl groups) increments as a result of the different alkyl substituents.

Many molecular classes are found in petroleum given the possible combinations in number and structure of rings and heteroatom (S, N, O) substitutions. They are all described by the SOL method, having different combinations of structural increments.

A complex petroleum mixture in SOL is thus composed of some 3,000-6,000 vectors (depending on boiling range), organized into about 150 molecular classes and their homologous series.

The relative amounts and carbon-number distributions of each molecular class can be determined experimentally using HDHA. HDHA examines, in detail, the molecular composition of petroleum and its refined products, which is necessary for understanding how crude composition and refining chemistry impact the composition and quality of refinery products. Mobil has developed an extensive HDHA library of crude oil and refinery product compositions.

Simulating processes

A characteristic feature of the catalytic and thermal chemistry of refining processes is that certain specific molecular rearrangements occur repeatedly on many different types of molecules. These particular increments correspond to those structural entities that are rearranged during reactions.

Employing a limited set of structural groups to describe thousands of components enables the use of a limited set of reaction rules to establish the complex reaction networks involving tens of thousands of reactions. A reaction rule identifies all molecules in the system that can undergo a certain type of structural rearrangement, and then generates the corresponding product. Each reaction rule consists of two parts, as follows:

1. Reactant-selection rule
2. Product-generation rule.

The product-generation rule converts each reactant's vector to a corresponding product vector. Examples of reaction rules for aromatic ring saturation, naphthenic ring opening, and dealkylation are given in Fig. 4 [98,437 bytes].

The aromatic-saturation rule determines if a molecule has the A4 ring required for the reaction, and then converts this ring to an N4 naphthenic ring through a simple mathematical operation. The reactant vector's A4 increment is decreased by one, and the N4 increment is increased by one to create the product vector. Molecules without an A4 ring would not be selected for this reaction.

Although hydrogen is also a reactant, it is not explicitly included in the formulation of the rule. Information on the hydrogen stoichiometry is automatically obtained from the difference in the hydrogen content of reactant and product molecules, and readily computed from the hydrogen content of the increments.

Applying a reaction rule to all molecules (or vectors) in the mixture generates a reaction class for the specific chemical transformation represented by the rule. A reaction class may have thousands of reactants and their corresponding products. Each reactant has the necessary increment for the rule.

SOL challenges

Application of all rules to all vectors generates the entire complex reaction network. Fig. 5 [62,578 bytes] illustrates how many potential products can be generated from just one molecule by the successive application of several rules. Similar networks are generated for each component.

Catalytic-refining processes have 20-40 reaction classes or rules, resulting in over 20,000-50,000 distinct chemical transformations for the model of the process chemistry. Computer programs using sorting procedures automatically use the network to construct differential rate equations for each component of the model.

Constructing complex reaction networks for thousands of components and their tens of thousands of reactions creates another challenge. The reactions require kinetic parameters, including rate constants, activation energies and adsorption constants on catalytic sites, and chemical thermodynamic properties.

Obviously, it is infeasible to conduct an experimental study of 50,000 isolated reactions. A first approximation assumes that all reactions in a class have the same kinetic parameters because they undergo the same intra-molecular transformation.

This assumption reduces the number of parameters for the entire network but generally proves insufficient. The molecular structure of the reactant within each reaction class influences the kinetic parameters for the reaction. That relationship between molecular structure and reactivity must be experimentally determined.

SOL application

All process models have the same common foundation-the SOL representation of molecular structure and composition. Each process is distinguished, though, by its unique set of reaction rules to represent the process chemistry and kinetics. The key catalytic reactions in lube hydrocracking, including aromatic ring saturation, denitrogenation, desulfurization, and cracking of low VI (viscosity index) components, result in high-quality base oils.

In the case of solvent processing, it is not chemical reaction networks that govern the transformation of petroleum to refined products, but rather, the phase-equilibrium thermodynamic properties of the individual molecular component. Solvent processing does not alter molecular structure. Rather, it distributes the molecular components between two liquid phases at equilibrium, for example, raffinate and extract, or wax and dewaxed oil.

Composition analyses (HDHA) of these phases from solvent extractions and dewaxing operations have led to new insights on the relationship between molecular structure and the thermodynamic properties of petroleum's components. These phase-equilibrium properties ultimately determine if a specific crude oil's unique molecular composition will result in a base oil with acceptable quality. Models of these solvent processes provide the operational guidance to optimize process conditions for base-oil quality.

For example, the SOL model can accurately predict compositions of raffinates by molecular species. In Fig. 6 [82,900 bytes], the model predicted 14 species against measured data at various furfural dosages. The model works well for all kinds of lube basestocks.

Determining composition

To determine a mixture property from a simulated or measured composition, structure-property correlations must be developed to estimate the contribution of the individual SOL molecular components. Mobil has developed a series of structure-property correlations based on the concepts of molecular class and homologous series (Fig. 7 [97,672 bytes]).

Using Mobil's SOL hydrocracking model as an example, the predicted yields of various boiling point cuts vs. measured values are shown in Fig. 8 [89,553 bytes]. The model was used to simulate the detailed composition. Boiling-point and specific-gravity correlations were applied to the simulated composition to obtain the predicted yields.

Key composition indicators for base-oil quality have been identified using the HDHA compositional analysis. Base-oil quality models have been developed from these indicators using advanced chemometric techniques and extensive product-testing experience. These quality indicators can be measured in the refinery for on-line monitoring and quality control, or predicted by SOL lube-process models in the assessment of a crude's suitability for base-oil production.

User interface

• Feed characterization. If time and budget are limited, the immediate need for results or unavailability of a sample requires the model to offer simpler alternatives. An extensive library of precharacterized feeds, techniques to mathematically construct new feeds through blending and tuning, connections to other process models, and a handle into the crude assay data base reduce the need for users to perform expensive and time-consuming analyses of their feeds while still providing accurate representation.
• Process flow. A process model is constructed from unit operations models, such as that of reactors, separators, distillation towers, and extraction columns. Each unit operation, based on the SOL modeling approach, has input and output streams with full HDHA-type composition detail. If desired, compositions of specific streams can be exported to commercially available simulators for detailed equipment analysis. Because a refinery may undergo small changes over time, the model provides users with an easy mechanism to configure new process flow diagrams. In addition to a large set of preconfigured diagrams, a simple drag-drop editor allows modeling of new units, modification of existing ones, and "what-if" studies.
• Product quality and yield predictions. Products are valued based on their bulk properties, such as viscosity, sulfur content, cold-flow properties, or boiling point. Composition-based property correlations provides users with properties they are accustomed to monitoring. Yields and conversion are usually based on boiling fractions, but may also be reported in terms of specific molecular classes or on an incremental basis for nitrogen and sulfur. The base-oil quality model operates directly on the detailed composition of model- output streams.

References

1. Jacob, S.M., Gross, B., Voltz, S.E., and Weekman Jr., V.W., "A lumping and reaction scheme for catalytic cracking," AIChE Journal, Vol. 22, No. 4, 1976, pp. 701-13.
2. Quann, R.J., and Jaffe, S.B., "Structure Oriented Lumping: describing the chemistry of complex hydrocarbon mixtures," I&EC Research, Vol. 31, 1992, pp. 2483-97.
3. Quann, R.J., and Jaffe, S.B., "Building useful models of complex reaction systems in petroleum refining," Chemical Engineering Science, Vol. 51, 1996, pp. 1615-35.

The Authors

Eugene Sanchez works for Mobil Technology Co. in the base-oil quality management/modeling group.

His responsibilities include defining the compositional boundaries of quality for mineral-based oils for Mobil's product quality standards. He joined Mobil in 1989 and initially worked in on-line infrared process analyzers. Sanchez holds a PhD in chemistry from the University of Washington, Seattle.